well position effects (edge and corner) effects in qPCR
Posted 13 October 2010 - 07:57 AM
I have been having some problems with well position effects in a qPCR assay using a BioRad iCycler iQ. Our machine is under service contract, so BioRad came (after much argument and delays) and replaced the lid, bulb and block. Still, I am seeing significant well position effects. BioRad says this is within the specifications of the machine and that there is nothing they can do about it.
Here are the specifics of the well position effect. Using BioRad's EvaGreen control assay with all samples exactly the same and carefully pippeted using a multichannel pippetman into bio-rad clear non-skirted plates with bio-rad optical seals I get remarkably similar effects across two different runs. Specifically:
Running a multivariate regression with Ct values as the dependent variable and corner and distance from nearest edge (on both axes) as independent variables I get the following:
-corner wells have a Ct value of ~0.25 higher
-for every well more separating a well from the edge, the Ct value goes up ~0.035 (e.g. well A1 has a score of 0, A2=1, B2=2 and C6=7.
The R^2 value is about 0.22
This means that going from an edge well to the inner-most wells = 7*0.035 or 0.245 Ct value difference.
Assuming a 100% efficiency to convert this Ct value to dilution factor (2^Ct=dilution factor) I have calculated that:
This means the most inner wells will tend to have an ~19% lower measured concentration than the most outer wells, and controlling for this effect, corner wells have an approximately 19% lower measured concentration.
Does this seem like a reasonable level of error to be acceptable and within specs? It seems like an awful lot to me--especially using BioRads control assay which is likely very well optimized.
My next step is to see if half-skirted plates, the optical compression pad and or the green ring thing going around the block helps to ameliorate this problem.
Posted 13 October 2010 - 09:48 AM
Posted 01 December 2010 - 08:59 PM
Based on linear regression predicting Cq with variables distance from closest two edges and dummy variable for corner:
Corner wells had a 0.49 higher Cq (initial conc differences of 41%)
each well farther away from the edge had -0.0108 lower Cq (difference from an edge well to most inner well of 14% initial concentration)
The regression model accounts for 17.5% of the variance (adjusted R2) - assay assembled with a Span8 robot, so pippeting error was minimized.
Too bad Bio-Rad seems to make their corner wells unusable and even in their more modern cyclers, can't get rid of their well-position effects.
Edited by dtae, 01 December 2010 - 09:04 PM.